Other primes thread

[Mini-Geek]

Originally Posted by mdettweiler

I think users will get banned at the top-5000 site if they submit too many composite primes. Not sure exactly how many's the limit though.

There's probably not a hard limit, it's probably more about whether they're obviously submitting things known not to be prime, like this one, whereas someone like us with many primes might get away with a couple of typos or CPU errors producing primes.

[MyDogBuster]

I was submitting a prime when the 3MM digit prime was being verified. I emailed Dr Caldwell because my prime was taking over 7 hours to verify. Dr Caldwell was none to happy about the 3MM digit prime because it was tying up the verification process on other submissions. He told me he had to put another machine on doing verifications because this prime had everything stymied. So, he knows about it.

[em99010pepe]

Who got an email from Dr Caldwell about Musatov false primes? He is asking for help to deal with this type of situation.

[gd_barnes]

Originally Posted by em99010pepe

Who got an email from Dr Caldwell about Musatov false primes? He is asking for help to deal with this type of situation.

I got that Email. I was also the one who notified him early this morning to delete Mustov, all of his accounts, and all of his obviously composite entries. He did just that.

I'm sure he'd like if I posted his Email here since he is asking for help, so here is a cut-and-paste of it:

Folks, I have a real problem. Martin M. Musatov is swamping my
primes.utm.edu site with false prime. His vandalism has really made
mess of the list. He first had confused a benchmark timer in proth.exe
with a primality test. Now he writes

From: Chris Caldwell
Subject: False primes
Sent: Jan 24, 2009 12:59 AM
> Why are you still submitting false primes?

From: Martin M. Musatov
m.mm@vzw.blackberry.net
Sorry, that was not intentional. I am trying to figure some things
out. Was 334928^262144+1 composite? Proth says no small factors.
Sent from my Verizon Wireless BlackBerry


I deleted his account and poof codes, he recreated them. I have blocked
the last IP addresses he was using, but he found a forth while I was
sleeping. I will delete all again in a few moments, and hack a line in
my code to block sending a password to his address, but I suspect he'll
get another soon. (And I have a meeting coming up, so it may b hours
rather than minutes.)

I could moderate all submissions to the list, but there would be long
slow periods as I was busy or traveling... folks like the quick
turn-a-round.

So I am asking here for any ideas. I know it is the wrong list, but
this group is often clever and even headed....

Chris

I suggested that he factor or sieve a little more before going into primality testing but that would not have helped the 2 submissions that took a long time recently.

To try to help out and allow an early deletion of the 1st large entry with no small factors, I sieved it to P=25G; no factors found. I then sieved the GFN that was taking all of last night to primality test to P=20*10^15 (20 quadrillion) over night on a very slow machine. Unfortunately, also no small factors. I stopped it after he went ahead and deleted the entry anyway.

I don't know what else to suggest to him than what he has already come up with above.

If anyone else has any ideas, I'm sure he would appreciate it.


Gary

[Mini-Geek]

Maybe automatically block any account with the name Musatov. (If there happens to be another Musatov that's a legitimate prime finder, he can contact you.)
Maybe some sort of smarter priority for which submissions are checked first could at least minimize the impact on verifying real primes.
Perhaps if somebody has more than one prime to verify or submits a large prime (over position 1000? I don't really know what would be a very reasonable amount), it would require Dr Caldwell's approval to verify. Or instead of simply 'more than one' or 'over position x', something that estimates the verification time for all submitted primes as 'over time x', i.e. if it'll take a while to verify all prime(s) make Caldwell have to okay it first; since this won't happen as often as every single prime, it wouldn't be unreasonable to force an extra wait for those before it shows up on the list. This would have to consider primes submitted separate times in addition to multiple primes at once.

[henryzz]

Originally Posted by em99010pepe

Who got an email from Dr Caldwell about Musatov false primes? He is asking for help to deal with this type of situation.

everyone on the primeform mailing list

[henryzz]

for some reason i cant submit my new prime
is this something to do with the composites problem?
only two have been submited today and they were really early today
edit: wierd tried again and the username and password box appeared this time

[Flatlander]

Maybe each prime submitted over a certain size has to be 'countersigned' by someone already on the list. The person already on the list gets an email, when they reply with their password or something the prime-checking starts. Sounds a bit clumsy though.

[MyDogBuster]

Didn't know where to post this so I'll put it here. Maybe we need a Mathematics for Dummies thread.

Question #1 for MFD: If twins are in the form of k*2^n-1 and k*2^n+1, then why don't we test all our prime found to see if it is a twin? Please remember geniuses, this is Mathematics for Dummies and I epitomize the title.

[gd_barnes]

Originally Posted by MyDogBuster

Didn't know where to post this so I'll put it here. Maybe we need a Mathematics for Dummies thread.

Question #1 for MFD: If twins are in the form of k*2^n-1 and k*2^n+1, then why don't we test all our prime found to see if it is a twin? Please remember geniuses, this is Mathematics for Dummies and I epitomize the title.

I had the same thought almost immediately after getting into prime searching. Therefore at that time, I checked Riesel primes for k<1200 for twins up to n=260K. (k=1200 is as far as the Proth list goes.)

It actually required no testing. I just cut-and-pasted lists of Riesel and Proth primes up to that limit into a spreadsheet and matched them up. Proth primes can be found at www.prothsearch.net. The largest twin that I found in doing this comparison was 915*2^11455+1 and -1. It had already been found, much to my chagrin. I found many more larger twins for higher k's. See below.

The chances of a prime found having a twin for n>260K are EXTREMELY remote. The largest twin at this point in history is at n=195K. It has a 10-digit k.

In order for k*2^n+1 and k*2^n-1 to be prime the k-value must be divisible by 3. So if you're up for comparing the Riesel and Proth lists, limit your comparison to k's divisible by 3 and n>260K.

With the few # of primes for n>260K and only 200 k's to check (600 odd k's divided by 3), you'll quickly see that far more k's are needed to have a reasonable shot at a twin.

Now...if you want to expand your search beyond the Prothsearch list, you could use the top-5000 site and test all Riesel's for any k-value that is divisible by 3.

I think Karsten has probably done something along these lines too. You might also check with him.

One more thing: I think you might find a couple of my web pages interesting:

Twins for k<100K and n<48K:
http://gbarnes017.googlepages.com/twins100K.htm

Twins for k<1M and n=10K-48K:
http://gbarnes017.googlepages.com/twins1M.htm

I have sieved files up to n=60K for these efforts. n=48K-52K is in midstream of testing on one of my quads. I stopped it about 7-8 months ago to focus on NPLB and CRUS.


Gary

[MyDogBuster]

Thanks Gary.

So basically, if I find a prime with a k divisible by 3, I could test for a twin but probably won't find one. Can't hurt though. I may even find one.